Rising AI costs force companies to cap spending, says Goldman Sachs

Uber used up its 2026 artificial intelligence (AI) budget in just four months, prompting it to restrict access and give developers priority over AI services for customers on its app, Goldman Sachs said in a report.

The bank describes Uber’s response as “strict gatekeeping on LLM usage”.

In its ‘Connecting the Dots’ report, the bank highlights how businesses are controlling AI bills by setting usage limits, removing tools that increase consumption and turning to smaller AI models.


“Many enterprises are currently capping AI spend, given elevated AI costs,” the report notes, drawing on examples from companies across technology, retail, telecom and financial services.

Much of this spending comes down to tokens, the pieces of text that AI models read and produce, which providers commonly use to calculate charges. Even small charges can add up when employees use these tools throughout the day, making it necessary to track both consumption and the value of the work.

Also read: As India buys more AI chips, the real test is to keep them running

Meta shut down an internal leaderboard that tracked AI usage because it encouraged “tokenmaxxing”, with the report linking the initiative to lead to an usage of 60 billion tokens in 30 days.

E-commerce giant Amazon also shut down an internal tool that encouraged token consumption and moved to allowances for individual teams, while Walmart moved away from “unlimited token usage” and introduced team allocations.

Microsoft removed access for some employees after facing high token costs per engineer and shifted its GitHub Copilot coding assistant to pricing based on credits, the bank’s compilation shows.

Predicting these bills has proved difficult, too. The report cites Cognizant’s warning that 85% of enterprises were missing their AI spending forecasts.

Companies are also changing the technology to lower the cost of completing each task, with some choosing smaller AI models designed for a narrower range of work.

Also read: AI costs spiral as firms lose spending control

“Cost of tokens can balloon,” the report notes in its summary on Capgemini, describing the challenge finance chiefs face when deciding how much money to set aside for IT. The IT services company is focusing on smaller AI models designed for specific types of work to improve efficiency and keep costs under control.

AT&T, whose average daily usage stood at eight billion tokens, changed the software that directs work to AI systems so it could use “smaller agents depending on the task”, a change that cut costs by 90%, according to the report.

Other businesses are focusing on what their spending achieves, with Zopa flagging unpredictable token costs and directing its AI budget towards improving resolution rates.

Meanwhile, Starbucks is building its own AI tools to reduce its software budget, while Spotify is weighing the cost of employees against spending more on AI tokens, the report said.